MBA HR to people analytics: attrition from a headcount register

The same artefact, before and after

The quickest way to see the switch is to compare what an HR generalist sends today with what a people analyst would send from the same data.

Every row on the right uses a document you already maintain. That is your advantage over a candidate who has never seen a real exit tracker: you know that "personal reasons" hides half the story and that a leaver's last working day and resignation date are different fields.

Before: the attrition number most decks get wrong

A common HR MIS habit is dividing the year's leavers by the closing headcount. That understates or overstates attrition depending on whether the company grew. The standard approach divides leavers in a period by the average headcount for that period. Below is a synthetic register and the calculation done both ways.

import pandas as pd

# Synthetic monthly headcount register for one department
reg = pd.DataFrame({
    "month":   ["2026-04", "2026-05", "2026-06", "2026-07", "2026-08", "2026-09"],
    "opening": [120, 124, 131, 135, 133, 142],
    "joiners": [8, 10, 9, 4, 11, 6],
    "leavers": [4, 3, 5, 6, 4, 7],
})
reg["closing"] = reg["opening"] + reg["joiners"] - reg["leavers"]

# Reconciliation check: this month's opening must equal last month's closing
mismatch = reg["opening"].iloc[1:].values != reg["closing"].iloc[:-1].values
print("Months failing reconciliation:", list(reg["month"].iloc[1:][mismatch]))

reg["avg_headcount"] = (reg["opening"] + reg["closing"]) / 2
reg["monthly_attrition_pct"] = (reg["leavers"] / reg["avg_headcount"] * 100).round(2)

total_leavers = reg["leavers"].sum()
avg_hc_period = reg["avg_headcount"].mean()
naive = total_leavers / reg["closing"].iloc[-1] * 100
proper = total_leavers / avg_hc_period * 100

print(reg[["month", "avg_headcount", "leavers", "monthly_attrition_pct"]])
print(f"Six-month attrition, leavers / closing headcount: {naive:.1f}%")
print(f"Six-month attrition, leavers / average headcount: {proper:.1f}%")

Run it and the reconciliation check flags September, where the synthetic opening does not match the previous closing. That is deliberate — a real register almost always has one, usually from a transfer between departments recorded on one side only. Finding and explaining it is worth more in an interview than the attrition percentage itself.

After: one question worth answering

A people analyst does not stop at the rate. Pick a single question your old manager actually asked, such as "are we losing people in their first six months?", and answer it from a synthetic exit tracker: tenure at exit bucketed into 0–6, 6–12, 12–24 and 24+ months, voluntary exits only, compared with the tenure mix of current staff. If early-tenure exits are over-represented relative to how many early-tenure staff you have, that points at onboarding, hiring source or role clarity — and each of those is testable with data you already have.

Write the project as a two-page memo plus the notebook: question, definition, data (declared synthetic), result, what it does not prove, what you would check next.

Where MBA HR training falls short for this work

• Formula-level Excel, then SQL. HR MBAs often use pivot tables well but have not joined an employee master to an exit table on employee ID. Learn joins early; people data is always split across tables.
• Definitions written down. Voluntary versus involuntary, regretted versus non-regretted, and whether contract staff count. Analysts get challenged on definitions before results.
• Small-number caution. A team of 12 losing 3 people is 25% attrition and also three individual stories. Say so.
• Privacy. Employee-level data is personal data. Aggregate before sharing, suppress tiny groups, and never use a real employer's register in a portfolio.
• A dashboard tool. Power BI or a similar tool, used for a small number of honest visuals.

HR analyst or HRIS analyst: picking the entry door

Look first at HR analyst, HR MIS analyst or people analytics associate roles, where your HR process knowledge is directly useful and the analysis is mostly descriptive. A broader data analyst title becomes realistic once your portfolio includes a non-HR project too. The free skill check will tell you which analyst skills your current resume does not yet show.

A sample answer for "Why leave core HR?"

"I spent three years as an HR business partner and owned the monthly headcount and exit reports. I realised the most useful part of my job was the one I did least carefully — the numbers. So I rebuilt our attrition report in Python on synthetic data, fixed the definition to use average headcount, and added a reconciliation check that caught a transfer recorded on one side only. I'd like to spend my time on that kind of work, and people analytics lets me keep the HR context I already have."

Notice what the answer avoids: it does not claim HR was boring, and it does not overstate the project. It names a definition, a check and a result.

If you want structured practice in SQL, Power BI and checking AI-assisted analysis before it reaches a manager, look at the Data Analyst programme.